Chinese AI startup DeepSeek has released a preview version of its highly anticipated Version 4 large language model, giving users early access to test its new capabilities. The Hangzhou-based company says the latest model delivers strong performance in agent-based tasks, knowledge processing, and inference compared to domestic competitors.
The V4 release marks another significant step in the escalating AI competition between Chinese and American tech companies. DeepSeek gained global attention in late 2024 with its free, open-source V3 model, followed by the R1 reasoning model in January 2025 that matched or beat leading AI systems from OpenAI and Google – reportedly at a fraction of the development cost.
Like its predecessor, DeepSeek-V4 follows the company’s open-source approach, allowing developers to download, modify, and run the code locally. This strategy contrasts sharply with the proprietary models from major US tech giants and has raised questions about the massive spending on AI infrastructure by companies like Microsoft and Google.
The new model comes in two versions – “pro” and “flash” – depending on size requirements. DeepSeek has optimized V4 for compatibility with popular development tools including Anthropic’s Claude Code and OpenClaw, making it easier for developers to integrate into existing workflows.
This release highlights the broader shift in AI development happening globally. When DeepSeek’s R1 model demonstrated that competitive AI systems could be built with less powerful chips and significantly lower budgets, it sent shockwaves through tech markets and forced investors to reconsider assumptions about AI infrastructure spending.
The timing is particularly significant as Chinese AI companies face growing restrictions on access to advanced semiconductors from US suppliers. DeepSeek’s ability to achieve strong performance with these constraints suggests alternative approaches to AI development that don’t rely on the most expensive hardware.
DeepSeek now faces intensifying competition within China’s booming AI sector. Major players are rushing to release new models:
- Alibaba has launched several updated AI systems this year
- ByteDance, the parent company of TikTok, continues expanding its AI capabilities
- Other Chinese startups are developing competing open-source alternatives
The company’s open-source strategy has broader implications for the global AI landscape. By making advanced AI capabilities freely available, DeepSeek is democratizing access to technology that was previously limited to well-funded tech giants. This approach could accelerate AI adoption across smaller companies and developing markets.
However, none of DeepSeek’s recent model updates have matched the market impact of the R1 release. The company will need to demonstrate significant improvements in V4 to maintain its momentum against both domestic competitors and international rivals.
The release also comes amid heightened tensions between the US and China over technology leadership. Reports suggest DeepSeek has been limiting access to its latest models for American engineers while providing early access to Chinese companies, further intensifying the technological divide between the two nations.
For the broader AI industry, DeepSeek’s continued success challenges the conventional wisdom that only massive budgets and the most advanced chips can produce competitive AI systems. This could reshape investment strategies and development approaches across the sector as companies look for more cost-effective paths to AI advancement.
